Patient-Level Predictors of Ebola Outcome

Descriptive Exploration of the 2014–2015 Simulated Outbreak

Shri Vishalini Rajaram

2025-05-12

-Background
-Patient Demographics
-Clinical Milestones
-Symptoms
-Viral Load
-Multi Comparisons
-Summary

Why Study Ebola Outcomes?

Background

  • Ebola virus disease (EVD) is a severe, often fatal illness caused by the Ebola virus.
  • With rapid clinical progression and high mortality, early detection and care may improve survival - yet detailed patterns remain unclear.

Dataset Summary

After pre-processing the dataset, we evaluated 5791 patients and 36 predictors between May 2014 and April 2015.

Table 1. Variable classification used in the analysis of Ebola outbreak patient data.
Variable Category Description
Patient Demographics Includes physical traits and hospital, which reflects treatment context.
Clinical Milestones Captures key dates and durations across the illness course.
Symptom Presentation Derived from symptoms column using string matching; totals as n_symptoms.
Viral Load Lower CT = higher viral load.
Clinical Outcome (Target) Recovery status of the patient: Recover, Death, or Unknown.

- Background
- Patient Demographics
- Clinical Milestones
- Symptoms
- Viral Load
- Multi Comparisons
- Summary
- Evaluation

Research question

Does the temporal changes of clinical milestones, symptom presentation, viral load, and patient demographics influence the clinical outcome of Ebola virus disease during the outbreak?

-Background
-Patient Demographics
-Clinical Milestones
-Symptoms
-Viral Load
-Multi Comparisons
-Summary

Distribution

How are patient demographics and clinical characteristics distributed?

A–B: Most patients were under 20; age distribution is right-skewed; C: Gender distribution was nearly equal among known cases; D: Many infection sources were unspecified (“other”); E–F: Most weights: 100–130 lb; most heights: 4.5–5.5 ft; G: Typical body temperature: 100–102°F; H: Port Hospital had the highest patient count.

-Background
-Patient Demographics
-Clinical Milestones
-Symptoms
-Viral Load
-Multi Comparisons
-Summary

Relationships

How does demographical changes affect outcome?

Left Recovery and death proportions remain relatively stable from ages 0–49, but the 50–69 group shows a visible increase in recovery and slight decline in death proportion.

Right Port Hospital handled the highest number of patients across all outcomes, especially deaths, suggesting it served as a major treatment center.

-Background
-Patient Demographics
-Clinical Milestones
-Symptoms
-Viral Load
-Multi Comparisons
-Summary

Relationships

Do overweight patients have higher risk of disease?

Weight increases with age as expected, but within each age group, clinical outcomes are not clearly separated by weight.

-Background
-Patient Demographics
-Clinical Milestones
-Symptoms
-Viral Load
-Multi Comparisons
-Summary

Distribution

Are there any patterns in the distribution of the clinical milestones?

Panels A–C show histograms of days between clinical events (A: infection, B: symptom onset, C: hospitalisation) and outcome.

-Background
-Patient Demographics
-Clinical Milestones
-Symptoms
-Viral Load
-Multi Comparisons
-Summary

Distribution

When was the peak time of the epidemic during 2014-2015 outbreak?

Monthly counts of cases for infection, symptom onset, and hospitalisation events. Negative values in A–C reflect likely data entry errors or same-day events.

-Background
-Patient Demographics
-Clinical Milestones
-Symptoms
-Viral Load
-Multi Comparisons
-Summary

Relationships

How did the hospitals adapt to the situation? Were there delays in treatment?

  • Military and Port Hospitals show stable and stratified patterns, indicating organized clinical pathways by outcome.
  • St. Mark’s and “Other” facilities exhibit greater month-to-month variation, suggesting either fewer cases or evolving protocols.
  • There is no strong evidence of systemic delay over time - rather, median durations remained stable or improved.

-Background
-Patient Demographics
-Clinical Milestones
-Symptoms
-Viral Load
-Multi Comparisons
-Summary

Relationships

Did clinical milestones vary among different age groups in turn affecting outcome?

  • Shorter time-to-outcome for deaths across all ages likely reflects severe presentations or delayed hospitalisation.
  • Longer recovery durations in older patients suggest sustained treatment efforts.
  • Variation in milestone intervals by age implies that age modified the timing of clinical trajectory, thereby influencing the likelihood of recovery or death.

-Background
-Patient Demographics
-Clinical Milestones
-Symptoms
-Viral Load
-Multi Comparisons
-Summary

Relationships

Average Time from Clinical Events to Outcome by Hospital and Outcome
Infection to Outcome Onset to Outcome Hospitalisation to Outcome
St. Mark's Maternity Hospital (SMMH)
Death 16.4 ± 8.3 7 ± 5.9 4.9 ± 5.7
Recover 25.3 ± 12.5 16.3 ± 10.9 14.4 ± 10.9
Unknown 22.9 ± 11.7 11.5 ± 9.8 9.2 ± 10
Port Hospital
Death 17.8 ± 9.9 7.4 ± 5.3 5.5 ± 5.3
Recover 27 ± 12.8 16.9 ± 11.2 14.8 ± 11.5
Unknown 20.8 ± 10.6 11.4 ± 7.6 9.1 ± 7.4
Other
Death 18.1 ± 8.9 7.8 ± 5.2 6.1 ± 5
Recover 25.5 ± 13.5 15.4 ± 11 13 ± 10.9
Unknown 19.9 ± 11.5 10.5 ± 8.6 8.5 ± 8.7
Military Hospital
Death 17.6 ± 9.9 7.4 ± 5 5.7 ± 5.3
Recover 26.9 ± 12.5 15.4 ± 9.7 12.9 ± 9.7
Unknown 20.4 ± 12.1 12.1 ± 9.7 9.8 ± 9.2

-Background
-Patient Demographics
-Clinical Milestones
-Symptoms
-Viral Load
-Multi Comparisons
-Summary

Distribution

Which was the highest recorded symptom among the patients?

Panels A–F show binary indicator counts for individual symptoms: fever, cough, vomiting, chills, aches, and absence of symptoms (“no symptoms”).

-Background
-Patient Demographics
-Clinical Milestones
-Symptoms
-Viral Load
-Multi Comparisons
-Summary

Distribution

How many symptoms were commonly reported?

Where there any left skewness with the viral load depicting higher disease?

-Background
-Patient Demographics
-Clinical Milestones
-Symptoms
-Viral Load
-Multi Comparisons
-Summary

Distribution

Were there any combination of symptoms prevalent in particular outcome?

Most patients with Fever+Cough or Fever+Cough+Vomit had a known outcome, with Death being the most frequent outcome in both combinations.

-Background
-Patient Demographics
-Clinical Milestones
-Symptoms
-Viral Load
-Multi Comparisons
-Summary

Distribution

Does the combination of reported symptoms and age category influence the clinical outcome?

  • For Fever+Cough, younger age groups (<30) consistently show higher recovery proportions, while older age groups have slightly increased death rates.

  • In Fever+Vomit and Vomit+Cough, recovery rates drop noticeably in middle-aged and older adults, with a corresponding rise in deaths, especially in the 50–69 group.

-Background
-Patient Demographics
-Clinical Milestones
-Symptoms
-Viral Load
-Multi Comparisons
-Summary

Risk profile

Can we find a risk profile for the children?
Age 0–9
Age 10–19
Death Recover Death Recover
Top Symptom Combo Fever+Cough+Vomit (43.7%) Fever+Cough (43.8%) Fever+Cough (43.9%) Fever+Cough (40.9%)
CT1 26.9 30.8 28.3 35.4
Weight2 51.5 51.1 98.9 99.5
Infection to Outcome3 16 (10–23) 26 (18.5–34) 16 (10–22) 24 (17–34)
Onset to Outcome3 7 (4–10) 14.5 (9–23) 7 (4–10) 14 (9–21)
Hospitalisation to Outcome3 5 (2–8) 12 (7–20) 5 (3–8) 11 (7–18)
Top Hospital Other Military Hospital Port Hospital Port Hospital
1 CT < 21 was the threshold and it indicates higher viral load across both age groups.
2 Weight thresholds: >80 lb for Age 0–9; >120 lb for Age 10–19. (Source: CDC, NCHS, 2024)
3 Median (IQR) reported for timing intervals in days.

-Background
-Patient Demographics
-Clinical Milestones
-Symptoms
-Viral Load
-Multi Comparisons
-Summary

Summary

What did we find?
  • Age matters: Younger patients had more deaths; older adults showed higher recovery.
  • Symptoms signal severity: Fever + Cough combos were common in fatal cases.
  • Weight \(\neq\) Risk: Heavier patients weren’t more likely to die after adjusting for age.
  • Faster deaths, slower recoveries: Time-to-outcome was shortest for deaths across all age groups.
  • Hospitals adapted over time: Port and Military hospitals showed stable care timelines.

THANK YOU